I'm going to come out of left field with unexpected optimism:
This is all speculation on my part, as a disclaimer. However, I think we're witnessing a Thomas Kuhn-style paradigm shift in RadOnc, and this research is the inevitable result of the cycle we've been in for the last 20 years.
What in the world am I talking about?
As I'm quite fond of pointing out, the Boomers have a deathgrip on RadOnc. These people finished residency from the late 1980s through the early 2000s. In their defense, the RadOnc technology of their time was totally different compared to what we have now.
As we all know, they doubled residency size and due to various factors, RadOnc was one of the most competitive specialties from about 2003/2004 - 2018/2019.
If you factor in the 5-year training lag, and the fact that people in power just don't give up power like, ever, there is an absolute army of some of the most brilliant people in American medicine just...stuck. That's a gross oversimplification of course, but a lot of institutional authority is locked up with people who have sat in those positions since George W Bush was president.
Since RadOnc is small, the cultural pressure is more intense - and cultural pressure in medicine is already intense at baseline. That means research in classic establishment institutions has to stick within narrow lanes considered "ok to study".
Those are:
1) Dose escalation (past peak popularity)
2) Omission (sustained popularity)
3) Hypofrac (peak popularity)
There's limited choices outside those three lanes. But you get things like:
1) Machine learning/AI (not a lot of people have the background to engage in that in RadOnc)
2) NCDB/SEER/other databases (very popular because low cost/barrier to entry, low impact)
3) Safety/QI
4) Environmental
5) Culture/DEI
The DEI space appears to be a path for the people who are in the DEI category themselves, and thus has limited generalizability.
If anyone thinks I'm crazy, go read the titles of the abstracts from this year's ASTRO conference, and the last few years. The "big" clinical trials are mostly hypofrac and omission. The residents are mostly engaged in database, QI, or environmental studies. The DEI group does DEI. There's an intense interest in AI but the pipeline to RadOnc is not historically the computer science crowd, so not a lot of people can get into that easily.
Simultaneously: the vast majority have soured on hypofrac and database work. Despite my jokes, there is a role for environmental studies - but for our specialty, for it to be the thing winning research awards and press releases...we've jumped the shark.
It's why we're seeing the interest in benign applications. Not just the low-dose stuff, but also cardiac ablation, SRS for various neurological disorders, etc.
The beginnings of the "crisis" in the paradigm shift is literally written in the literature. Look at the people who published the COVID lung trial. Then, look at the people who authored the "this is bad" anti-COVID lung paper. That's the divide.
Further, everyone keeps talking about FLASH. From a radbio perspective, if you think FLASH (and GRID/LATTICE) are "real" or at the very least plausible, existing theories about radiation biology are insufficient to explain those results. It will require adoption of other theories and models which are currently not showing up on the ABR exams.
So we're at a crossroads. Radiation will remain important in certain cancer applications, as long as cancer exists, and the field can't totally die while that remains true. The "powers that be" will have to leave somehow, either through retirement or their own death, and the army of early career people will be able to fill the vacuum.
None of this is going to be fun. But the 2030s and beyond is probably going to look very different than 2023.
Med students should still look elsewhere, though.